Biophysical Correlates of Cognition Among Depressed and Nondepressed Type 2 Diabetic Patients

RESEARCH DESIGN AND METHODS—

An ongoing study on diabetes and depression is being conducted at UCLA Medical Center. Depressed and nondepressed patients were under care provided at UCLA clinics affiliated with departments of internal medicine and endocrinology. Subjects were screened with the Structured Clinical Interview for DSM-IV (SCID) (4), the Mini Mental Status Examination (MMSE), the American Heart Association's Stroke Risk Prediction Chart (CVRF) (5), and the Cumulative Illness Rating Scale (CIRS) (6). Participants completed an electrocardiogram and a standard battery of laboratory tests. The study protocol was approved by the UCLA Institutional Review Board. Please see Elderkin-Thompson et al. (7) and Kumar et al. (8) for complete procedures.

Neuropsychological battery

The neuropsychological battery included tests of working memory (Letter-Number Sequences), language (Controlled Oral Word Association, or FAS), simple attention and processing (Stroop 1 and 2, Trailmaking A, Digit Symbol Substitution), procedural learning (Wisconsin Card Sorting Test), executive function (Matrix Reasoning, Trailmaking B, Stroop Interference), visuospatial conceptualization (Block Designs), and declarative learning and recall (California Verbal Learning Test, trials 1–5, short and delayed recall). Cronbach's α = 0.90 for the composite global scale indicated good reliability.

Methods

MRI was performed with a 1.5 Tesla scanner (Signa, Lx echospeed plus 9.1; General Electric Medical System, Milwaukee, WI) using a transmit/receive quadrature head coil. Imaging was performed in an axial plane using fast-spin echo T2 and MT T1 weighted sequences. Details on MRI acquisition and MT methods have been previously reported (9,10). In MT imaging, an off-resonance radio frequency pulse is used to minimize the exchange of protons between bound water (bound to macromolecular proteins) and free water compartments in the brain (11,12). The resulting image is a subtraction of the image with the RF pulse from the original one without the pulse.

Statistical plan

Demographic and clinical variables were analyzed with ANOVAs and post hoc contrast tests (Table 1). MT ratios across groups were compared using CIRS as a covariate because diagnostic groups differed in their symptom counts. CIRS, CVRF, and short-form (SF)-36 were highly intercorrelated, so CIRS was selected as the most representative of general health status. Cognitive scores were standardized, with higher scores representing better performance and tested for homogeneity (Cronbachs α = 0.89). The global cognitive scale was correlated with MT ratios after CIRS adjustment. Brain regions that correlated with the cognitive index were examined by plotting the MT ratio by the cognitive index.

RESULTS—

Diagnostic groups differed on global cognition, with both diabetic groups performing significantly below the comparison group and depressed performing significantly below nondepressed diabetic patients. MT ratios from the caudate differed across diagnostic groups and formed the same step decline of means observed in the cognitive index.

Correlations

MT ratios of caudate and putamen correlated with the cognitive index after adjustment for CIRS total symptom count (caudate, r [59] = 0.36, P = 0.006; putamen, r [59] = 0.44, P = 0.001). Correlations remained significant after Bonferroni's correction (α = 0.0125). Correlation coefficients for the remaining regions were small to trivial.

In the subgroup analysis, healthy comparison individuals did not demonstrate a relationship between cognition and caudate MT ratios (r [28] = 0.24, P = 0.23). Diabetic patients as a group did show a relationship (r [31] = 0.37, P = 0.04), but there was no significant difference between the correlations of control subjects and patients (Fisher's z difference = 0.52, P = 0.60). When the diabetic group was further subdivided into depressed and nondepressed patients, only the depressed diabetic subjects showed an association (r [13] = 0.56, P = 0.05), but this also was not significantly different from healthy comparison subjects (z difference = 1.04, P = 0.30).

In the putamen, control subjects demonstrated no linear relationship between MT ratios and cognition (r [28] = 0.08, P = 0.70). Diabetic patients showed a relationship (r [31] = 0.59, P = 0.001), which was significantly different from healthy control subjects (Fisher's z difference = 2.17, P = 0.03). In the patient subgroups, both nondepressed and depressed diabetic patients showed an association (r [18] = 0.50, P = 0.04; r [13] = 0.66, P = 0.01, respectively), although neither correlation was significantly different from the healthy comparison group (z = 1.44, P = 0.15; z = 1.90, P > 0.05).

Scatterplots (not provided) demonstrated the different patterns of association for the two regions. In the caudate, data points of diagnostic groups showed some segregation with three apparent clusters of scores, consistent with significant mean differences. The healthy comparison group was clustered primarily above the MT and cognition means, depressed diabetic subjects were largely below both means, and nondepressed diabetic subjects were grouped at the means. In the putamen, the data points from the groups were interspersed with no group mean differences.

CONCLUSIONS—

As hypothesized, global cognitive performance correlated with caudate MT ratios. The finding that the MT-cognition correlation was stronger among patients then among healthy comparison subjects, and stronger among depressed diabetic patients than nondepressed diabetic patients, suggests a threshold pattern in which the MT effect is modest but notable when only diabetes is present, but it becomes exacerbated when both diabetes and depression are present. MT ratios from the putamen correlated with global cognition, but the means did not differ between groups, so there was no obvious disease effect. Although interpretation of low MT ratios in white and gray matter are not yet understood, physiological abnormalities in cell membranes and proteins together with neuronal and synaptic loss are currently posited as explanations for differences.

A methodological caution results from the small sample size and low power of the analysis. Differences between groups may become significant with a larger sample, and the role of the putamen deserves further study. In conclusion, our results, although preliminary, suggest that one mechanism underlying mild cognitive decline in diabetes may lie in abnormalities of the protein biochemistry of the caudate.